I argued (commenting in Alan Rhoda's weblog here) that
(1) It is possible that there are unicorns.
does not imply
(2) There are unicorns
Alan disputes this, arguing that (1) can be formalised
(3) (∃w)(Ww & (∃x)(xEw & Ux))
But how do we translate this formalisation? If it translates 'it is possible that there are unicorns' we still can't infer that there are unicorns. The 'that' clause acts, as it is designed to, to protect us from any inference as to the truth of the statement embedded in the clause (just as it protects us from inferring 'there are unicorns' from 'Jack thinks that there are unicorns'. If, on the other hand, we translate it as
(3') There is a possible-world, and there are unicorns, such that the unicorns are in that world
then this does logically imply there are unicorns. But that is because the translation strips out the 'that' clause. Which begs the question. If we are allowed to translate 'it is possible there are unicorns', which does not imply there are unicorns, by a sentence that contains 'there are unicorns' as a logical component, and which does for that reason imply there are unicorns, then Alan has pulled off the trick. But are we allowed to?
Whatever is said in formal logic, seems deep.